Energy-saving driving control system and method for rail vehicle

Abstract

This invention discloses an energy-saving driving control system and method for rail vehicles. The system includes: a perception layer that collects train operating status and traction parameters in real time; a generalized resistance potential field module that constructs and updates a generalized resistance potential field characterizing the resistance throughout the entire process using a dual-channel fusion architecture; a traction efficiency establishment module that calculates the dynamic traction efficiency coefficient based on static efficiency and the rate of change of traction force; a rolling optimization decision module that uses discrete speed points as decision variables and aims to minimize the equivalent traction energy consumption to generate the optimal speed and traction force sequence online; and a driving suggestion output module that converts the optimal sequence into executable instructions with physical cause identifiers. This invention achieves high-precision adaptive modeling of resistance, integrates static and dynamic losses of the traction system into optimization, considers both global optimization and onboard real-time calculation, and generates interpretable driving suggestions, thereby improving energy-saving effects and human-machine collaboration trust.

Description

Technical Field

[0001] This invention relates to the field of rail transit technology, and in particular to an energy-saving driving control system and method for rail vehicles. Background Technology

[0002] As a high-capacity, low-energy-consumption land transportation mode, energy optimization has always been one of the core driving forces for technological development in railway transportation. Train traction energy consumption accounts for a considerable proportion of the total operating cost of railways, and developing advanced energy-saving driving technologies has significant economic and social benefits. However, existing train energy-saving control technologies still have many shortcomings in practical applications.

[0003] Traditional traction calculations generally rely on fixed empirical formulas with coefficients that are statistical averages. These formulas fail to reflect changes in actual factors such as train formation, equipment status, ambient wind, and wheel-rail adhesion, resulting in model deviations of ±30% to ±40%. This leads to distorted outputs from optimization algorithms based on these formulas, hindering the realization of theoretical energy-saving potential. Regarding optimization algorithms, global optimization methods such as dynamic programming are computationally complex and require large storage. They typically only allow offline calculation of speed curves and cannot handle real-time disturbances such as temporary speed limits and signal delays. Simplifying the model to reduce computational power weakens the optimization effect. At the optimization objective level, existing optimization objectives often aim to minimize traction mechanical work, neglecting the static efficiency (MAP) of the traction motor and converter, as well as dynamic losses such as switching losses and iron losses caused by sudden changes in traction force. This may lead to increased actual power consumption. Existing driving suggestion systems primarily output speed curves or simple prompts, making it difficult for drivers to understand the physical reasons behind their decisions. This results in low trust levels, difficulties in human-machine collaboration, and a significant reduction in energy-saving effects.

[0004] In summary, current energy-saving train driving technology faces a series of dilemmas, including inaccurate models, insufficient computing power, shallow optimization, and lack of trust between humans and machines. A systematic solution is urgently needed, ranging from the reconstruction of the underlying physical model to the top-level human-machine interaction design. Summary of the Invention

[0005] To address the above problems, this invention provides an energy-saving driving control system and method for rail vehicles. By constructing a generalized resistance potential field and a dynamic traction efficiency coefficient, it achieves high-precision, low-computational-burden, and interpretable real-time energy-saving driving control.

[0006] In a first aspect, the present invention provides an energy-saving driving control system for rail vehicles, comprising: The sensing layer is used to collect real-time train operating status, track environment, and dynamic parameters of the traction system. The generalized drag potential field module is used to construct and update in real time the generalized drag potential field that characterizes the drag of the train throughout the entire operation process based on the data collected by the sensing layer. The traction efficiency establishment module is used to calculate the dynamic traction efficiency coefficient based on the static efficiency of the traction system and the rate of change of traction force. The rolling optimization decision module is used to generate the optimal speed and traction force sequence online under dynamic constraints, with discrete speed points within the look-ahead window as decision variables and minimizing the traction energy consumption equivalent as the objective. The dynamic constraints are constructed based on real-time resistance information, and the traction energy consumption equivalent is calculated based on the dynamic traction efficiency coefficient. The driving suggestion output module is used to convert the optimal speed and traction sequence into executable driving commands with physical drive factor identifiers, and output them through the human-machine interface.

[0007] Furthermore, the generalized drag potential field module adopts a dual-channel fusion architecture, including: The direct observation channel, based on the principle of energy conservation, uses real-time back-propagation of potential field observations using net wheel circumference traction force, train mass, acceleration, and train speed; The parameterized synthesis channel constructs a parameterized potential field model based on first principles, which includes gravitational potential energy, curve-additional potential, aerodynamic potential, and adaptive basic drag potential. An adaptive extended Kalman filter is used to fuse the potential field observations with the parameterized potential field model, learn online, and output the optimal field estimate and the basic drag parameter set.

[0008] This invention achieves real-time, high-precision estimation of the generalized drag potential field through a dual-channel fusion architecture. The direct observation channel, based on the principle of energy conservation, uses sensor data to infer potential field changes in real time. The parameterized synthesis channel constructs a potential field model incorporating gravity, curves, aerodynamics, and adaptive basic drag based on physical laws. The adaptive extended Kalman filter then fuses the two, enabling online parameter learning and optimal state estimation. This architecture ensures the reliability of the physical model while possessing strong adaptability to changes in the actual operating environment, significantly improving the accuracy and robustness of drag modeling.

[0009] Furthermore, the functional form of the adaptive basic resistance potential is: ; in, This is the overall scaling factor, which controls the overall magnitude of the basic resistance potential; The contribution coefficient to low-speed frictional resistance; This is a coefficient representing the rate at which frictional resistance saturates with speed; The basic drag component is proportional to speed; t represents time; arctan is the arctangent function; This is the fundamental resistance potential; s represents the train's position; v represents the instantaneous speed of the train.

[0010] A parameterized model incorporating the arctangent function is used to describe the fundamental drag potential. Compared to the traditional quadratic polynomial model, this model can more accurately fit the high static friction characteristics in the low-speed region and the air drag saturation characteristics in the high-speed region. This functional form has stronger nonlinear expressive power. Combined with online learning of time-varying parameters, the model can adapt to different train formations, equipment wear, and environmental changes, thereby significantly improving the accuracy and generalization ability of fundamental drag modeling.

[0011] Furthermore, the dynamic traction efficiency coefficient is defined as: ; in, The static efficiency of the traction system; The dynamic traction efficiency coefficient; This is the switching loss penalty factor; This is the penalty coefficient for incremental iron loss; For wheel circumference traction; This represents the rate of change of traction force around the wheel. This represents the motor speed.

[0012] The dynamic traction efficiency coefficient proposed in this invention not only considers the static efficiency of the traction system but also introduces a penalty mechanism for the increase in switching losses and iron losses caused by sudden changes in traction force. This elevates the optimization objective from the traditional "minimizing mechanical work" to "minimizing overall electrical energy consumption," guiding the system to automatically generate a smooth traction force change curve, avoiding inefficient operating areas, thereby reducing energy consumption, electrical and mechanical stress, and extending the service life of key components of the traction system.

[0013] Furthermore, the rolling optimization decision module specifically includes: Discretization unit is used to divide the look-ahead window into several equidistant segments, with the velocity at the end of each segment as the decision variable; The constraint construction unit, based on the equivalent kinetic energy theorem, uses the potential field gradient of the generalized resistance potential field to calculate the average generalized resistance of each segment, and establishes the relationship between traction force and velocity between adjacent velocity nodes. The objective function construction unit takes the sum of the traction energy consumption equivalents of each segment as the objective function. The traction energy consumption equivalents are calculated from the traction force, speed, time step and dynamic traction efficiency coefficient. The solution unit is optimized by using a sequential quadratic programming algorithm combined with a hot-start technique to solve the optimal velocity sequence that satisfies the constraints of speed limit, time, traction force and acceleration online.

[0014] By discretizing the look-ahead window into speed node decision variables and constructing concise dynamic constraints using the equivalent kinetic energy theorem, the dimensionality of the optimization problem is greatly reduced. Employing a sequential quadratic programming algorithm and warm-start technology, the complex optimal control problem can be solved in real time on the onboard computing platform, enabling rapid response to dynamic disturbances such as temporary speed limits and schedule adjustments, significantly improving the practicality and adaptability of the energy-saving driving strategy.

[0015] Furthermore, the driving suggestion output module includes: The instruction segmentation unit is used to divide the driving process into traction, cruise, coasting or braking intervals based on the optimal traction force sequence. The physical cause identification unit is used to analyze the dominant energy factors in each interval and generate labels for physical causes. The Natural Language Transformation Unit is used to combine instruction segments and tags into text suggestions.

[0016] By segmenting instructions, identifying physical causes, and converting them into natural language, complex optimization outputs are transformed into driving suggestions that are easy for drivers to understand and execute. Drivers can not only see what to do, but also understand why to do it, greatly improving the system's explainability and trustworthiness, promoting human-machine collaboration, and ensuring that energy-saving strategies are effectively implemented in actual operation.

[0017] Furthermore, the system also includes a fault diagnosis and degradation control module, which is configured as follows: Real-time monitoring of sensor data and dual-channel observation residuals; switching the system to the corresponding operating state based on the severity of the fault, including: Dual-channel fusion, full-function optimization; When a non-critical sensor fails, increase the corresponding process noise covariance; When a critical sensor fails, shut down the direct observation channel and use the parameterized synthesis channel and simplified optimization strategy. In case of a severe fault, optimization suggestions will be disabled, and only basic circuit and safety prompts will be output.

[0018] This invention incorporates a multi-level fault diagnosis and degradation control mechanism, capable of real-time monitoring of sensor data and dual-channel observation residuals, intelligently determining the fault type and severity, and automatically switching to the corresponding degradation operation mode. From full-function optimization to simplified strategies, and then to a safety fallback mode, the system can maintain basic functions and driving safety under different fault conditions, greatly improving the system's engineering reliability and field adaptability.

[0019] Secondly, the present invention provides an energy-saving driving control method for rail vehicles, comprising: S1 collects real-time train operating status, track environment and dynamic parameters of traction system through the sensing layer; S2 utilizes a dual-channel fusion architecture to construct and update the generalized resistance potential field in real time, and outputs the optimal field estimate and the basic resistance parameter set. S3, calculate the dynamic traction efficiency coefficient based on the static efficiency of the traction system and the rate of change of traction force; S4 uses discrete velocity points within the look-ahead window as decision variables and aims to minimize the equivalent traction energy consumption to generate the optimal velocity and traction force sequence online under dynamic constraints. S5 converts the optimal speed and traction sequence into executable driving commands with physical drive factor identifiers and outputs them through the human-machine interface. S6 monitors system health in real time and executes corresponding degradation strategies based on the fault level.

[0020] Compared with existing technologies, the beneficial effects of this invention are as follows: By constructing a generalized resistance potential field through a dual-channel fusion architecture, high-precision real-time modeling and adaptive learning of train running resistance are achieved, breaking through the bottleneck of low accuracy in traditional fixed empirical formulas; simultaneously, a dynamic traction efficiency coefficient is introduced, unifying the static efficiency of the traction system with dynamic losses such as switching losses and iron losses caused by traction force changes into the optimization objective, achieving a leap from "minimizing mechanical work" to "optimizing comprehensive electrical energy consumption"; a rolling optimization strategy using discrete speed points as decision variables, combined with sequential quadratic programming and hot-start technology, balances global optimality with onboard real-time computing capabilities; and interpretable driving suggestions are generated through physical cause identification and natural language conversion, significantly improving human-machine collaboration trust; furthermore, multi-level fault diagnosis and degradation control mechanisms ensure the system's reliability and driving safety under various abnormal operating conditions. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this drawing or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this drawing. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0022] Figure 1 This is a structural framework diagram of the system of the present invention; Figure 2 This is a schematic diagram comparing the static efficiency and dynamic comprehensive performance of the traction system at its operating point in the embodiment. Figure 3 This is a schematic diagram showing the locations of the high-efficiency and low-efficiency operating points in the embodiment; Figure 4 This is a graph showing the penalty effect of the traction force change rate on the overall performance in the embodiment. Figure 5 This is a state transition diagram for the multi-level fault diagnosis, fault tolerance, and degradation strategy of the system of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments provided by this invention without inventive effort are within the scope of protection of this invention.

[0024] This invention provides an energy-saving driving control system for rail vehicles, such as... Figure 1 As shown, it specifically includes: The sensing layer is used to collect real-time train operating status, track environment, and dynamic parameters of the traction system, specifically including: Traction sensing: A high-frequency response variable torque sensor is installed on the output shaft of each traction motor. The signal is transmitted through a non-contact measurement system, and at the same time, the motor phase current, bus voltage and switching frequency data are read from the traction converter control unit. Motion attitude perception: An inertial measurement unit (IMU) is installed and integrated with a BeiDou dual-frequency high-precision receiver to achieve centimeter-level position and velocity measurement through a tightly coupled integrated navigation algorithm; Environmental perception: An ultrasonic three-dimensional anemometer is installed in the center of the top of the locomotive driver's cab to measure the wind speed, wind direction and turbulence intensity at a distance of 3 meters above the rail surface; Status awareness: Read train formation information, vehicle identification number, gross weight, and load data from the train control and management system; Maps and Knowledge Base: The onboard storage device is pre-loaded with an electronic map of the route, which includes gradient, curve radius, speed limit, tunnel location and cross-sectional dimensions, typical wind zone information, and precise coordinates of stations and signals.

[0025] The collected raw data is aligned to the same time base point through hardware timestamps or software interpolation. The torque signal is filtered by a 10Hz low-pass Butterworth filter to remove high-frequency mechanical noise. The IMU acceleration signal is tilt compensated and the coordinate system is rotated to ensure that its axis is strictly parallel to the longitudinal direction of the rail.

[0026] The generalized drag potential field module is used to construct and update in real time a generalized drag potential field characterizing the drag of the train throughout its operation, based on data collected by the sensing layer. The physical meaning of this potential field is: the energy per unit mass required to move a unit mass train from a reference point to its current position s with a speed v, in order to overcome all types of resistance (including dissipative forces related to speed) from the starting point to the current position.

[0027] Furthermore, the generalized drag potential field module adopts a dual-channel fusion architecture, including: The direct observation channel, based on the principle of energy conservation, uses real-time back-propagation potential field observations using net wheel circumference traction force, train mass, acceleration, and train speed.

[0028] Specifically, calculate the net wheel circumference traction. : ; in, This refers to the number of traction motors on the locomotive. To account for the safety factor of wheel adhesion, a factor of 0.92 is used; The effective wheel diameter at the current moment; Net torque, , The instantaneous shaft torque read from the torque sensor. This is a transmission system efficiency model, which is a three-dimensional numerical table, i.e. Obtained through bench testing, for example, in terms of torque. Rotation speed oil temperature At that time, look up the table to get .

[0029] Calculate the true acceleration : ; in, For gravity; θ is the slope angle. , It is the arctangent function. This is the precise slope value for the current location, retrieved from the electronic map of the route; Acceleration measured by the IMU.

[0030] Online train mass estimation : when and At that time, the online estimate of train mass is updated using the recursive least squares method with a forgetting factor (FFRLS), and the update formula is as follows: ; ; ; in, Kalman gain (updated gain); To estimate the error covariance; I is the identity matrix; For real acceleration, for The transpose of the matrix; Theoretically, the entire net force after deducting the drag estimated by the current model from the traction force should be used to generate acceleration. , This is the current estimated resistance value; The forgetting factor is set to 0.995.

[0031] Calculate the potential field observations: Calculate the power consumed instantaneously to overcome non-inertial drag. : ; According to the definition of a potential field, the total derivative of the potential field along the train trajectory is... The relationship is: ; Numerical integration of the above relationship using the trapezoidal rule yields the sequence of observed potential fields along the train trajectory: ; in, The speed at the current moment; , These are the time derivatives of the observed potential fields at the current and previous moments, respectively; The sampling time interval; , The observation sequences for the current time and the previous time are respectively, U obs The initial value is 0.

[0032] The parameterized synthesis channel constructs a parameterized potential field model based on first principles, which includes gravitational potential energy, curve-additional potential, aerodynamic potential, and adaptive basic drag potential.

[0033] Specifically, parameterized potential field model The structural design is as follows: ; in, This is the change in gravitational potential energy. ; Add potential to the curve. ; K c This is the curve resistance conversion factor. The radius of the curve; The aerodynamic potential, whose integral reflects the influence of the spatiotemporal variations of the wind field, ; air density, The drag coefficient varies with the crosswind angle. β change, For the train's frontal area, This represents the wind speed projection along the track direction (positive for headwind). It is an adaptive function describing the fundamental resistance potential, and its form is: , This is the overall scaling factor, controlling the overall magnitude of the basic drag potential. It varies depending on changes in train conditions (such as load and bearing lubrication status). It will adapt and adjust accordingly; The contribution coefficient to low-speed frictional resistance; This is a coefficient representing the rate at which frictional resistance saturates with speed; The basic drag component is proportional to speed; A set of time-varying parameters is used to make the model adapt to a specific train.

[0034] An adaptive extended Kalman filter is used to fuse the potential field observations with the parameterized potential field model, learn online, and output the optimal field estimate and the basic drag parameter set.

[0035] Specifically, the output of the two channels and The input is fed into an adaptive extended Kalman filter (AEKF) with parameter learning, and the state vector X of the AEKF is defined as: ; Where U represents the generalized resistance potential field.

[0036] The extended Kalman filter dynamically and optimally corrects the state estimate by comparing real-time observations with predicted values, while simultaneously learning the basic resistance characteristic parameters of the current train.

[0037] Specifically, state prediction includes: ; ; ; ; in, , , These represent the predicted position, velocity, and generalized drag potential field, respectively. , , These represent the observation location, velocity, and generalized drag potential field, respectively. , To predict time-varying parameters; , For observing time-varying parameters; , These are the partial derivatives of the generalized drag potential field with respect to position and velocity, respectively, and are approximate values ​​calculated from the state at the previous moment. It is acceleration.

[0038] Calculate the predicted state covariance matrix : ; Where F is the state transition Jacobian matrix. Let F be the transpose of F; Q is the process noise covariance matrix; Let be the observation state covariance matrix.

[0039] The observation update process is as follows: Obtain observation vector ,in, The position measured by the GNSS positioning system at the current moment. The speed measured by the speed sensor, These are the potential field observations output from the direct observation channel; Calculate the observation prediction vector ,in, The potential field prediction value is calculated from the predicted state through a parameterized synthesis channel; Calculate the observation residuals ; Calculate Kalman gain Where R is the observation noise covariance and H is the observation Jacobian matrix; Corrected state estimation ; Update covariance .

[0040] from Extracting the most dominant field and through The optimal resistance estimate is obtained by performing finite difference on position s. ,Right now ,in This represents the displacement increment.

[0041] The traction efficiency establishment module is used to calculate the dynamic traction efficiency coefficient based on the static efficiency of the traction system and the rate of change of traction force.

[0042] This invention recognizes the nonlinear characteristics of traction system energy consumption and proposes a dynamic traction efficiency coefficient. Specifically defined as: ; in, The static efficiency of the traction system was obtained through bench tests. The dynamic traction efficiency coefficient; This is the switching loss penalty factor. The term is used to penalize the additional switching losses and current stress caused by sudden changes in traction force, and its directional insensitivity is reflected in absolute value form; This is the penalty coefficient for incremental iron loss. The term is used to penalize the increase in motor iron loss caused by rapid changes in traction force, and is proportional to the square of the rate of change, which physically corresponds to eddy current loss. For wheel circumference traction; This represents the motor speed.

[0043] This comprehensively reflects the energy efficiency level of the traction system at a specific operating point and under specific conditions. Incorporating this into the optimization objective means that the algorithm will not only avoid the low static efficiency zone, but also automatically generate a smooth traction force variation curve, thereby reducing electrical and mechanical stress and extending equipment life while saving energy.

[0044] like Figure 2-4 As shown, static efficiency is compared using the following four operating points. With dynamic comprehensive efficiency The difference, and show the effect of the rate of change of traction force (dF / dt) on The effect of the punishment: 1) The highest static efficiency point: Location: =180kN, =100 rad / s; at this time, =0.9000, =0.5542; 2) Recommended high-efficiency work locations: Location: =180kN, =120 rad / s; at this time, =0.6449, =0.3971; 3) Inefficient work areas to avoid: Location: =40kN, =60 rad / s; at this time, =0.5000, =0.3307; Dynamic penalties resulted in a 33.9% decrease in effectiveness; 4) Rate of change of traction force Impact ( =150kN, =100rad / s): dF / dt = 5kN / s: =0.5058, efficiency decreased by 21.6%; dF / dt = 10 kN / s: =0.4030, efficiency decreased by 37.5%; dF / dt = 20 kN / s: =0.2687, efficiency decreased by 58.3%; dF / dt = 30 kN / s: =0.1897, efficiency decreased by 70.6%.

[0045] The rolling optimization decision module is used to generate the optimal speed and traction force sequence online under dynamic constraints, with discrete speed points within the look-ahead window as decision variables and minimizing the traction energy consumption equivalent as the objective. The dynamic constraints are constructed based on real-time resistance information, and the traction energy consumption equivalent is calculated based on the dynamic traction efficiency coefficient.

[0046] This invention proposes an efficient, vehicle-mounted rolling optimization framework. Its core is to solve a highly simplified but physically meaningful optimization problem within a finite look-ahead window (e.g., 2-5 kilometers ahead). The algorithm is triggered at a frequency of 1Hz, optimizing the look-ahead length each time. The distance is 3000 meters. The specific steps are as follows: The look-ahead window is discretized into N segments, and the velocity at the end of each segment is selected as the decision variable. There are a total of N-1 variables. v N fixed.

[0047] Specifically, all discrete points within the look-ahead interval are obtained from the electronic map of the route. slope curvature Speed ​​limit .

[0048] The current train status is: , ,in, The starting position (i.e., the current position of the train) ), The starting speed (i.e., the current train speed) ); Terminal speed constraint: Set the speed at the look-ahead endpoint as the safe target speed, which is usually the speed limit at that point (if the endpoint is a station, set it to 0). Time budget : Calculate the arrival time at the target destination based on the running chart The remaining planned time.

[0049] Based on the equivalent kinetic energy theorem, for the k-th segment (from... arrive ), and derive the discrete-form dynamic relationship: ; in, , These are the speeds at the beginning and end of the segment, respectively. This is the length of the segment; This represents the average traction force of that segment; The average generalized resistance of this segment is determined by the current most advantageous field. The gradient is obtained by averaging over the interval, i.e. .

[0050] The optimization objective is to minimize the energy consumption equivalent, with the sum of the traction energy consumption equivalents of each segment as the objective function J. Specifically, the objective function J is: ; in, This represents the average speed of that segment; The time it takes to pass through this period; This represents the average motor speed for that segment; Estimate the rate of change of traction force between adjacent segments; The objective function drives the algorithm to find the operating strategy that maximizes the overall efficiency of the traction system.

[0051] The solution unit is optimized by using a sequential quadratic programming algorithm combined with a hot-start technique to solve the optimal velocity sequence that satisfies the constraints of speed limit, time, traction force and acceleration online.

[0052] Specifically, the constraints include: speed limit constraints. Terminal speed constraint Remaining runtime constraints Furthermore, the amplitudes of traction force and acceleration are both less than the corresponding preset thresholds.

[0053] The above problem is a medium-sized planning problem with nonlinear objectives and constraints. This invention uses the Sequential Quadratic Programming (SQP) method to solve it, and the specific process is as follows: The objective function and constraints are combined into a Lagrangian function. : ; in, The original objective function; , For Lagrange multipliers; This is an inequality constraint function; This is the equality constraint function.

[0054] At the current iteration point The original problem can be approximated as a quadratic programming subproblem: The objective function is approximately a quadratic form: ; in, Indicating the search direction; It is the Lagrangian function at the current iteration point The Hessian matrix at that location; For the objective function J at the current iteration point The gradient at that point.

[0055] The constraint is approximately linear: ; in, For inequality constraint functions in gradient at; for Inequality constraint values ​​at; For equality constraint functions in gradient at; for The equality constraint value at the location.

[0056] Solve the above quadratic programming problem using the activity set method or the interior point method to obtain the search direction. ; along Perform a one-dimensional search in the direction (satisfying the Armijo condition) to determine the step size. Update decision variables ; like If the value is less than the tolerance and the constraint violation is sufficiently small, then stop the iteration and output the optimal solution. .

[0057] Shift the solution from the previous optimization cycle forward. We use a set of location points as the initial guesses for this optimization. Due to the continuity of train status and track, the solutions of adjacent cycles are usually very close, which can significantly reduce the number of iterations and enable the SQP algorithm to converge within 10 to 20 steps. The time taken for a single optimization is usually 100 to 300 milliseconds, which fully meets the requirements of onboard real-time computing.

[0058] The driving suggestion output module is used to convert the optimal speed and traction sequence into executable driving commands with physical drive factor identifiers, and output them through the human-machine interface.

[0059] Specifically, based on the optimal velocity sequence Calculate the corresponding average traction force Traverse each segment: like Then it is marked as the traction section, in which The dead zone threshold for traction force; like If so, it is marked as the braking section; Otherwise, mark it as a lazy segment.

[0060] Merge adjacent segments of the same type to form continuous instruction intervals, and calculate key parameters for each instruction segment, including suggested operation (such as "traction to 85km / h"), starting / ending position, and suggested traction / braking gear.

[0061] The system analyzes the main driving forces behind this instruction segment and assigns a physical label to each instruction segment. The specific method is as follows: Calculate the potential energy change for each segment. and kinetic energy change And identify the main contributors; like Then it is marked as potential energy conversion, where This is the change in gravitational potential energy; like If the value is greater than the corresponding threshold and is strongly correlated with the time constraint, it is marked as kinetic energy management; If the average traction force of this segment is less than the corresponding traction force threshold but is a dynamic traction efficiency coefficient If it is significantly improved compared to the previous stage, it is marked as performance optimization; like If the proportion is abnormally high, it is marked as environmental adaptation, among which... This represents the change in aerodynamic potential energy.

[0062] By combining instruction segments with physical cause labels, natural language suggestions are generated using the format of "action + goal + cause". For example, a suggestion might be to accelerate to 85 km / h to conserve kinetic energy for an uphill climb.

[0063] The driver's cab dashboard displays both graphical forward speed curves and text command cards on a separate screen. This presentation method connects control theory with the driver's intuition, greatly improving the credibility of the suggestion system and the efficiency of human-machine collaboration.

[0064] To ensure high system reliability, this invention designs a multi-level fault diagnosis and fault-tolerant processing mechanism, and automatically switches operating modes according to the severity of the fault, such as... Figure 5 As shown, the specific steps include: Sensor-level monitoring: Real-time verification of the rationality, range, rate of change, and signal continuity of data from each sensor; Channel consistency diagnosis: Monitor the observed deviation sequence in the adaptive extended Kalman filter (AEKF). If the deviation of a certain channel (such as the directly observed channel) continues to exceed the statistical confidence interval, it is determined that the data of that channel may be faulty.

[0065] Depending on the severity of the fault, the system will switch to the corresponding operating state, including: Mode A (Normal): AEKF uses both the direct observation channel and the parametric synthesis channel simultaneously, performing rolling optimization full run; Mode B (Level 1 Degradation): When a non-core sensor (such as an anemometer) malfunction is detected, the process noise covariance of AEKF is automatically increased, the weight of fault data in the fusion is reduced, and the system performance is slightly degraded. Mode C (Level 2 Degradation): When a critical sensor (such as torque) fails, only the parametric synthesis channel is used to calculate the optimal field and net wheel circumference traction. By working backward from the motor current and nominal efficiency, the optimization algorithm is switched to a simplified version that only considers static efficiency. Mode D (Safety Net Plan): In case of a serious fault, the system disables optimization suggestions and only displays basic line information and safe speed prompts.

[0066] The present invention also provides an energy-saving driving control system for rail vehicles, comprising the following steps: S1 collects real-time train operating status, track environment and dynamic parameters of traction system through the sensing layer; S2 utilizes a dual-channel fusion architecture to construct and update the generalized resistance potential field in real time, and outputs the optimal field estimate and the basic resistance parameter set. S3, calculate the dynamic traction efficiency coefficient based on the static efficiency of the traction system and the rate of change of traction force; S4 uses discrete velocity points within the look-ahead window as decision variables and aims to minimize the equivalent traction energy consumption to generate the optimal velocity and traction force sequence online under dynamic constraints. S5 converts the optimal speed and traction sequence into executable driving commands with physical drive factor identifiers and outputs them through the human-machine interface.

[0067] S6 monitors system health in real time and executes corresponding degradation strategies based on the fault level.

[0068] It should be noted that the present invention is not limited to the above-described embodiments. The above embodiments are merely examples, and any embodiments that have the same structure and perform the same effects as the technical concept within the scope of the present invention are included within the scope of the present invention. Furthermore, various modifications that can be conceived by those skilled in the art to the embodiments, and other ways of constructing by combining some of the constituent elements of the embodiments, without departing from the spirit of the present invention, are also included within the scope of the present invention.

Claims

1. An energy-saving driving control system for rail vehicles, characterized in that, include: The sensing layer is used to collect real-time train operating status, track environment, and dynamic parameters of the traction system. The generalized drag potential field module is used to construct and update in real time the generalized drag potential field that characterizes the drag of the train throughout the entire operation process based on the data collected by the sensing layer. The traction efficiency establishment module is used to calculate the dynamic traction efficiency coefficient based on the static efficiency of the traction system and the rate of change of traction force. The rolling optimization decision module is used to generate the optimal speed and traction force sequence online under dynamic constraints, with discrete speed points within the look-ahead window as decision variables and minimizing the traction energy consumption equivalent as the objective. The dynamic constraints are constructed based on real-time resistance information, and the traction energy consumption equivalent is calculated based on the dynamic traction efficiency coefficient. The driving suggestion output module is used to convert the optimal speed and traction sequence into executable driving commands with physical drive factor identifiers, and output them through the human-machine interface.

2. The energy-saving driving control system for rail vehicles according to claim 1, characterized in that, The generalized resistance potential field module adopts a dual-channel fusion architecture, including: The direct observation channel, based on the principle of energy conservation, uses real-time back-propagation of potential field observations using net wheel circumference traction force, train mass, acceleration, and train speed; The parameterized synthesis channel constructs a parameterized potential field model based on first principles, which includes gravitational potential energy, curve-additional potential, aerodynamic potential, and adaptive basic drag potential. An adaptive extended Kalman filter is used to fuse the potential field observations with the parameterized potential field model, learn online, and output the optimal field estimate and the basic drag parameter set.

3. The energy-saving driving control system for rail vehicles according to claim 2, characterized in that, The functional form of the adaptive basic resistance potential is: ； in, This is the overall scaling factor, which controls the overall magnitude of the basic resistance potential; The contribution coefficient to low-speed frictional resistance; This is a coefficient representing the rate at which frictional resistance saturates with speed; The basic drag component is proportional to speed; t represents time; arctan is the arctangent function; This is the fundamental resistance potential; s represents the train's position; v represents the instantaneous speed of the train.

4. The energy-saving driving control system for rail vehicles according to claim 1, characterized in that, The dynamic traction efficiency coefficient is defined as: ； in, The static efficiency of the traction system; The dynamic traction efficiency coefficient; This is the switching loss penalty factor; This is the penalty coefficient for incremental iron loss; For wheel circumference traction; This represents the rate of change of traction force around the wheel. This represents the motor speed.

5. The energy-saving driving control system for rail vehicles according to claim 1, characterized in that, The rolling optimization decision module specifically includes: Discretization unit is used to divide the look-ahead window into several equidistant segments, with the velocity at the end of each segment as the decision variable; The constraint construction unit, based on the equivalent kinetic energy theorem, uses the potential field gradient of the generalized resistance potential field to calculate the average generalized resistance of each segment, and establishes the relationship between traction force and velocity between adjacent velocity nodes. The objective function construction unit takes the sum of the traction energy consumption equivalents of each segment as the objective function. The traction energy consumption equivalents are calculated from the traction force, speed, time step and dynamic traction efficiency coefficient. The solution unit is optimized by using a sequential quadratic programming algorithm combined with a hot-start technique to solve the optimal velocity sequence that satisfies the constraints of speed limit, time, traction force and acceleration online.

6. The energy-saving driving control system for rail vehicles according to claim 1, characterized in that, The driving suggestion output module includes: The instruction segmentation unit is used to divide the driving process into traction, cruise, coasting or braking intervals based on the optimal traction force sequence. The physical cause identification unit is used to analyze the dominant energy factors in each interval and generate labels for physical causes. The Natural Language Transformation Unit is used to combine instruction segments and tags into text suggestions.

7. The energy-saving driving control system for rail vehicles according to claim 1, characterized in that, It also includes a fault diagnosis and degradation control module, which is configured as follows: Real-time monitoring of sensor data and dual-channel observation residuals; switching the system to the corresponding operating state based on the severity of the fault, including: Dual-channel fusion, full-function optimization; When a non-critical sensor fails, increase the corresponding process noise covariance; When a critical sensor fails, shut down the direct observation channel and use the parameterized synthesis channel and simplified optimization strategy. In case of a severe fault, optimization suggestions will be disabled, and only basic circuit and safety prompts will be output.

8. A method for energy-saving driving control of rail vehicles, characterized in that, include: S1 collects real-time train operating status, track environment and dynamic parameters of traction system through the sensing layer; S2 utilizes a dual-channel fusion architecture to construct and update the generalized resistance potential field in real time, and outputs the optimal field estimate and the basic resistance parameter set. S3, calculate the dynamic traction efficiency coefficient based on the static efficiency of the traction system and the rate of change of traction force; S4 uses discrete velocity points within the look-ahead window as decision variables and aims to minimize the equivalent traction energy consumption to generate the optimal velocity and traction force sequence online under dynamic constraints. S5 converts the optimal speed and traction sequence into executable driving commands with physical drive factor identifiers and outputs them through the human-machine interface. S6 monitors system health in real time and executes corresponding degradation strategies based on the fault level.

Images